Welcome to Fractal Forums

Maybe I need to start with simple zoom sequences, but I can't find a decent movie making tutorial for doing zoom movies in KF2.

I want to make a zoom movie, preferably 1280x720 with at least 4x4 antialias.

If I option to store zoom out sequence, I get a series of Jpeg images with corresponding kfr files. How do I make this into a movie? Also I need anti-aliasing turned on but I can't find the option to do AA renders. I normally just render big and scale down using bilinear filtering if I want to render stuff. Also I have witnessed the image size reset if the image is bigger than the window.

Sorry for being a newb, but I'm only ever rendered videos using Fractal Extreme.

The area I would like to zoom into is deep on the edge of the main cartorid. Everything looks like sandstone and the iteration data is in the millions at a depth of just 190 zooms. I can see why people don't normally zoom into the super duper dense areas but bands do become visible with sufficient anti aliasing and very long color cycles. The dense banding somewhat resembles sandstone and I think it will make a neat zoom movie.

EDIT: NEvermind, I forgot the movie maker is in a separate download.  :headbatting:

I still need to know how to do anti-aliased movies... :angry:

There is no supersampling antialias function in Kalles Fraktaler, antialias is only achieved with picture scaling.
The reason is that the frames are stored as grids of iteration values, instead of bitmaps with RGB values.
The reason for that is because averaging iterations doesn't make a good result, and it is not possible to do color cycling, at least the way I want, or change the palette with immediately effect etc.
I believe Mandel machine have the same kind of structure.

Anyway, you can render the frames up to 3840x1440 and if you make the result movie at 1280x720 you get at least 3x3 antialias.
Further the next frame is squeezed and pasted into to center, so 25% of the view will have 6x6 antialias.
And the 25% on the center will have yet the next frame on it, 12x12 antialias.
And so on, I usually combine 6 frames in my movies. The center, where you usually focus your eyes, is very antialiased.

The reason for the size limit of 3840x1440 is that KFMM is 32-bit. But if you use a lower number of combined frames you might be able to use bigger frames.
The KFB file format is open and others have done their own movie assembler programs. Seryzone on this forum.

hi stardust!
keep in mind that as soon as you use "rotation" in the moviemaker the resolution will be halved.
so to make a rotating full had movie (without alias at the outer edgeds) in full hd you will have to choose the max resolution of 3840*2160.

I've tried doling "post"-2*2-antiaalias in virtualdub but it didn't really improve too much in my opinion.

Looking forward to see movies from you!

Cheers

...and I suggest you don't store the frames as jpeg.
You can examine the frames with the programs (Examine zoom sequence in KF and previews in KFMM)

Yes, and this is especially needed for lots of locations in the Burning Ship. Some areas like this one are pretty hard to navigate through in low-res without AA. Just see the difference:

(http://i.imgur.com/3kaAjLM.jpg)

960px - no AA

Do we see dust? Grass?


...


(http://i.imgur.com/rAP3kKf.jpg)

7680px scaled back to 960px

Ahh, shapes pop out of the soup


Btw, very nice that stardust4ever is going to do deep zoom movies. Already looking forward to the results!   :o

Thanks for the info guys. So, I don't need to save the frames as Jpeg?

I found an area within the Mandelbrot set. I zoomed into the edge of the main cartoid. The depth of the final mini is only 192 zooms but the iteration density is rediculous. The max iteration depth is 5,000,000 and it seems even Mandel Machine is sluggish (took 5 minutes to render the frame at 1280x960 4x4 AA, 1000000 iterations on an 8-core @4.2Ghz). Here's a screenshot from near the end of the zoom. Scaled to 800x600 for size. Using UltraFractal palette, iterations scaled to -310.

Progress is slow go. 33 hours in. I'm having a lot of issues right now.

I configured KF to render each frame at 3840x2160. Some of the frames are being saved unfinished. Also the first non-black frame got repeated 3 times. Either way, there are multiple frames where the image is incompletly rendered with perturbation glitches and outside pixels at the edge of the image being blockier than the inner pixels. It's as if for several of the frames so far, the software is advancing to the next frame for unknown reason before it is fully rendered. The incomplete frames do not have iteration glitches corrected while the complete ones do. I have a feeling this will cause artifacts when the video is rendered. Is there some way to manually rerender only the frames that contain glitches? Not sure why it is doing this. I have linked some images and have attached the zipped source file.

Here are the first ten images working backwards from 1E57:

[problem solved - links deleted]

Is the final zoom video gonna be FUBAR?

AMD bulldozer (8 cores) @4.2Ghz (OC)
Brand new RAM: 16Gb DDR3 Raedon @1867Ghz (8Gb x2, dual channel)
Windows 7 64-bit Professional

Hi stardust

I suspect the jpeg function doesn't work so well on movie rendering.
I also suspect the jpegs are stored as black images if KF is minimized.
I also suspect the problem is only in the jpeg images, not the KFB files.

You can examine and correct frames with the function "File->Examine zoom sequence".

I maybe should remove the option to create jpegs on zoom sequences, because it only causes problems, and I never use it myself....

No!

You should make an image sequence the only available output for movies   O0

Less overhead and more control over the final product. jpg is too lossy though, perhaps a lossless image format like png would be ideal.

I know - those are not likely requests, but somebody had to say it   :evil1:

When KF is minimized or otherwise out of focus (other desktop screen under linux), it seem to keep the half-way progress of the last image while KF was still in focus. So now I just use the Linux (Openbox) options to make the window stickey ( = visible on each desktop). My experience is that the window itself does NOT have to be active/focussed, but it must be on the current active desktop/screen.

I can confirm that this problem is only with the JPEG creation, the KFB map files are still good even though the images are messed up. So if the jpegs are the only problem, just ignore and use the Key Frame Movie Maker on the map files.

I'll take your word for it. I quit last night at 4am and when I cyecked the PC today at 1pm, there were nearly a dozen or so Jpegs that were identical copies of the last screen I veiwed before I re-minimized the render. I generally keep render apps minimized as it's less likely that a stray keystroke or mouse click will disrupt the program. I always minimize to the desktop when I leave the PC. In case the cat walks across the keyboard, etc... I haven't checked it this evening but it what you're saying is accurate, there will be a lot of duplicates of the unfinished render I saw at 1pm. As long as the KFR iteration data is valid, I'm good. IMO the frame renderer needs some work. Lots of options but no help file or documentation to describe what the options do. I'll need to scale the output from 3840 to 1280 otherwise the output video will be ginormous. Each KFR file is ~63Mb, but I don't care as I have 100s or gigabytes free on my harddrive. I tupically use 1280x720p X264 high profile constant quantizer q17 either Slower or Placebo to render videos. Fractal data is generally harder/longer to compress and takes up much larger file size than live video, but that's okay by me. My DSL uplink is 200kb, so it takes a long time to upload. As usual, Youtube compression ruins the videos.

All controls (I hope) are at least described on http://www.chillheimer.de/kallesfraktaler
The movie tutorial is somewhat outdated though, since with Pauldelbrot's glitch detection examine and correction of frames should not be necessary.

I hope you succeed, you should perhaps have started with a simpler zoom.
But I'll help you :)
(today the KF support department will drink a lot of beers though, with friends from the country side visiting the capital ;))

Yeah it takes time getting used to new software. Had it been a simple shallow zoom movie, I'd have stuck with Fractal Extreme. I can see why nobody zooms into this type of area within the Mset (Edge of main cartoid). Iterations are so dense they have a space-filling effect where the empty space forms inverse dendrites instead of the other way around.

Perturbation based rendering still breaks down if you can find a shallow area with excessive iteration depth. The speedup over traditional rendering is not as great factoring in millions of iterations per pixel which still need to be calculated despite the lower precision. 1E57 is the minibrot and I'm two days in although frames are getting faster.

I have the color palette (4 colors) set to mimic UltraFractal with a very sparse repeat rate of 40x. Zoomed out the Mset appears black inside and out because black is the zero index color. Images are still grainy even at 3x3 antialias and I have a feeling each frame jump during the video will be visible since resolution increases twofold changes along the edge of every frame.

Perhaps a better way to design zoom movies would be one long image of arbitrary height (height is equal to zoom level with further down being higher magnification, each horizontal line drawing a circle around the center) using polar coordinates with a logarithmic sliding scale. The Mset is a single 2D plane so a zoom movie could in theory be represented by a single 2D polar bitmap rather than a series of 2D rectangular images. Temporal quality will increase approaching the center of the video frame but all frames will have the smooth image quality with no "jumps" along the edges of each frame. Assuming the circumference of the cylindrical tube is sufficiently large in pixel count, ultra smooth zoom videos would result.

maybe someday in the future when computers have a trillion petabytes of memory  :)

Rendering finished earlier tonight. Now trying to generate a single AVI under 2Gb. My last attempt at Q20 was 3 minutes, 11 sec and 2.44Gb. VLC plays it with a popup warning about rebuilding the index, but nothing else will touch it. There is also a lot of image grain which causes the constant quantizer bitrate to increase. YouTube will most likely puke all over it but if I can find a host for the avi files (Mega, etc...), that would be nice.

Also is there some way to change the text color? Magenta kind of clashes with my chosen color palette.

I just think polar coordinates would be a better rendering method than a rectangular frame per every zoom. Fractal Extreme and Kalles Fraktaler output videos occasionally you can see the quality increase along the edge of each key frame. Using polar coordinates around the zoom center would create a perfectly smooth zoom video with infinite opportunities for rotation. Key frames can be generated periodically to prevent an antialias bottleneck around the centroid.

Got a nice minty AVI file: 3'19", 1280x720, x264, High Profile, Film preset, Constant Quatizer =23, Very Slow, 1.85 Gbytes.

Funny thing, you can't even see the zoomed out Mandelbrot in the early seconds due to the slow color banding and black zero index.

I'll get it uploaded to Youtube. Hopefully YT doesn't destroy the quality too much but I know it will. Fingers crossed... :hurt:

Coming to the "Rate my Video" forum soon... :dink:

Wow, Youtube really puked on this one. I normally use H264, High Profile, CQ 17. To save a proper AVI, I had to go up to CQ 23 just to get the 3'19" video to fit under 2Gb. Youtube compresses it to less than one twelfth it's original file size and the results are not pretty...

http://www.fractalforums.com/movies-showcase-%28rate-my-movie%29/sandstone-zoom-into-the-edge-of-the-large-cardioid/

http://www.youtube.com/watch?v=_WfjrLupX_Y

 :vomit:

I think it is still excellent!
If you would upload the original to Mega, I would definitely download it :)

Just an FYI, the bit rate is crazy high so you'll need a fairly robust CPU to view the movie without hiccups. That shouldn't be an issue for most on these forums but I doubt it will play successfully on any mobile devices.

I signed up for a Mega.co.nz account today. Currently uploading the zoom movie, but it is slow go @50kb/s. About 10 hours left... Surprisingly it's not my DSL that's limiting the uplink speed (I typically get 200-250kb/s uplink). I assume there is some sort of file permissions setting to make the file links public. I would never trust the "cloud" for private stuff. Buy a pair of external drives for backup and go nuts...

http://www.rawstory.com/rs/2014/09/01/nude-photographs-of-a-list-celebrities-stolen-from-cloud-service-and-posted-online/
 :embarrass:

https://mega.co.nz/#!K1kSjB7I!wjd2wExJPdji0EepUnbpbIMSR1rHykHPkzFtSR4shy0 (https://mega.co.nz/#!K1kSjB7I!wjd2wExJPdji0EepUnbpbIMSR1rHykHPkzFtSR4shy0)
 ;D

Excellent!

The movie looks corrupted at start on my laptop, but from e007 it works perfect.

Unfortunately the frame edges are visible on the densest passages.
Maybe I should consider blending the pixels instead of just pasting the next squeezed frames on top of the others.
And should I try to differentiate the blend ratio by measuring how far the pixels are from the center...
FX is blending the frames, right?

The youtube remix was just ... no comment.


The H264, High Profile, CQ 17 is WAY better! Pleasure watching. I like the 4-way saddle points, which are a nice blend of smooth geometrics and the more grainy fractal stuff. The shapes found between e46 and e48 create some kind of optical illusion. The zooming makes my brain think it also rotates ... which it doesn't.  :whistle: :nerd:

:hoppingmad: Why hasn't some color cycling been added to this zoom ... ?!?

X264 CQ=23 (lower number = higher quality and file size). I normally use CQ 17 for fractal vids as a small amount of tearing results from smooth gradients at higher compression levels but this zoom has very little smooth gradients aside from the opening seconds and CQ23 still preserves the details. Try rewinding the zoom as it loads if it's tearing on start as it might still be buffering the video when playback starts. You really do need a fast CPU (>2Ghz) to play it without glitching. My 2006 MiniMac (Upgraded to OSX 10.6 Core 2 Duo @2.0Ghz) can't play it without glitching and neither does my fiance's 2010 Lenovo laptop (Windows 7 32-bit, 2.2Ghz Core 2 duo). In hindsight, I probably should have checked the "reduced CPU" option which decreases the algorithm efficiency. I primarily use VLC Media Player btw.

As for encoding, I used 3480x2160 frames for 3x3 antialias. Somebody warned me not to go over 4k resolutions, plus it took 3 days to render even at moderate zoom level. Really this area needs around 6x6 antialias to really look smooth, and even then it would still have a sandstone like texture.

Rectangular frames are a problem when rendering Fratal videos regardless of the way they are blended because there is always a 2x increase in definition across the edges of frames. I can often see two such jumps if I freeze-frame the AVI: Equivalent to just a hair over 1.5x1.5 near the edge,  3x3 in the interior, and 6x6 in the center. Fractal Extreme stores zoom frames as bitmaps (8-bit for standard zooms; 24-bit for AA zooms) rather than raw iteration data, but the effect is still similar. FX frames tend to be slightly blurry/pixellated towards the edges, KF is slightly more noisy.

I need to make a post about this somewhere, but perhaps polar coordinates would be a better strategy for fractal zoom videos. I came up with the idea some time back while watching some of my own FX video renders of the ABS variants (burning ship, etc) and noticed the edges of frames were sometimes visible. Mandelbrot is a flat 2D plane. Think of the zoom path as a long cylindrical tube rather than a progression of still images. Pixels get smaller at an exponential rate as you zoom down. Keyframing would only be necessary on the terminous of the journey (to prevent a circular black hole), and to prevent boghing the CPU by needlessly antialiasing millions of pixels at the center. No sudden changes in definition as with rectangular frames, but a smooth increase in definition towards the center. Also with polar coordinates, the fractal could be rotated endlessly without loss of fidelity. The quality would be defined by specifying the circomference and depth per zoom in pixels. For a clean antialiased picture, you would want the circomference of the zoom tube to be larger than the perimeter of the output video frame.

As for colors, I used my own simple 4-color interpretation of the default "Ultrafractal" color palette because I liked the way it looked in Mandel Machine:
(0,0,0)
(255,128,0)
(255,255,255)
(0,128,255)
repeat infinitely, 40x scaling...

I haven't yet experimented with waves or color cycling yet, but I plan on holding onto the iteration frame data for a while so I assume tgey can be rerendered with a new palette. Each of the 193 frames is 63 megabytes, so the files are way too big to upload. Honestly I just prefer simply constant velocity zoom sequences as I don't really have the patience to bother with keyframing.